Chromeless phase shift lithography (CPL) has been investigated for many years as a possible single-mask resolution enhancement technique for lines/spaces in semiconductor devices. For positive resists, it is particularly well suited to the patterning of semi-isolated narrow lines but not to dense line/spaces or contacts. However, with significant mask design effort and added mask complexity, contacts and semi-dense line/spaces have been successfully patterned. Like other phase shifting techniques such as alternating PSM lithography CPL can provide significantly better aerial image contrast compared to binary masks; unlike alternating PSM lithography however, it is a single mask single exposure technique avoiding many of the dual-reticle concerns such as throughput, mask layout, and reticle to reticle overlay.
CPL uses phase edges between 0 and 180° phase shift regions on the mask to pattern lines along the phase edges. This is possible without chrome because destructive interference of light diffracted from regions immediately on either side of the phase edge result in an aerial image minimum at the wafer corresponding to the phase edge with excellent contrast if it is isolated enough. With just one phase edge defining lines, it would be impossible to pattern arbitrary layouts without a second mask to clear unwanted phase edges. CPL allows one to avoid using a second mask by patterning narrow lines with two closely spaced parallel phase edges that cannot be resolved. The combined aerial image of the two parallel phase edges is still a deep single minimum which patterns as one line but now the “line” on the reticle (mask) can be drawn just as it would with chrome, wherein the chrome is replaced by a phase shifted region. However, this only works for lines that are not wide; if the phase shifted line becomes too wide, i.e. the two phase edges of the line move too far apart, then they become individually resolvable and will pattern as two parallel lines. If the phase shifted line is too narrow, the aerial image contrast gets worse very quickly as the phase shifted region become smaller and looks more like a uniform piece of quartz. These two cliffs constrain the size of phase shift lines to a relatively tight range of small widths.
These effects are illustrated in the aerial image diagram of FIG. 1, and the schematic diagrams of FIGS. 2A and 2B. FIG. 1 corresponds to an aerial image intensity distribution simulation of a CPL reticle that is illuminated with 193 nm light using on-axis quadrupole illumination (0.1 sigma poles at 0.7 sigma radii), and projected using a 0.68 NA (numeric aperture) projection lens. The ideal case corresponds to a 0.1 μm separation, which produces a deep single minimum. As the separation width increases, the aerial image results in a pair of minimums being produced, as shown by the 0.2 μm and 0.5 μm separation curves. For example, a separation of 0.5 μm would result in two lines being resolved. This of course is undesired. As a result, wider lines are typically patterned using a binary (i.e., chrome-patterned) reticle.
The results of the foregoing phenomenon are shown schematically in FIGS. 2A and 2B. FIG. 2A depicts an aerial image 200 produced by a CPL mask 202 that includes a narrow phase-shifting feature comprising a mesa 204 having a width W1. After the resist is exposed, developed, and washed, and a metal layer is deposited over a substrate 206, and single line 208 is formed. The single line 208 has a width W1′ that substantially matches the width W1 of narrow mesa 204 (or is otherwise proportional thereto for projection systems that employ magnification). As illustrated in FIG. 2B, when short wavelength light is directed at a CPL mask 210 including a wide mesa 212 phase-shifting feature with a width W2, the resulting aerial image 214 includes two narrow peaks rather than one wide peak. As a result, two narrow lines 216 and 218 are formed instead of a desired wide single line.
Under conventional practices, this wide line width/feature limitation of CPL is addressed by providing a mask that employs both CPL features and chrome features. The CPL phase shift features are used to produce narrow features, while chrome patterns are added to the CPL mask to produce large area features such as wide lines and pads on the semiconductor substrate. In this instance, the chrome is used to block light rather than phase shift the light, as is well-known in the art. One disadvantage of this approach is that the mask making process becomes more complex. Extra lithographic and etch steps in the mask making process are required to make both the chrome features and the CPL features. In addition the chrome and CPL patterns need to be precisely aligned.